The GammaVar model was developed for fitting a SINGLE gamma variate shaped component in the data -- that is, there is only one 'stimulus'. This thing was developed for modeling the response to an injected drug -- up and down (or down and up if the amplitude is negative!).
This model has 4 parameters, and as always in NL, you need to set the ranges over which program searches for these carefully or you will not get decent results. These parameters are
1) time delay of response from start of images == t0
2) amplitude factor == k
3) rise exponent == r
4) decay constant == b
The formula for the response model is
f(t) = 0 for t <= t0
f(t) = k * (t-t0)^r * exp(-(t-t0)/b) for t > t0
The reason I give the formulas, although I know many psychologists abhor such things, is that they are the only unambiguous way of being clear on what is being calculated.
You especially need to be careful when using this model to set the limits on t0 (parameter #1) reasonably.
If you are looking for a gamma variate DECONVOLUTION model, then you need one of the other functions.